Integrating Genetic Data and Electronic Medical Records to Reassess Variant Pathogenicity in the Taiwanese Han...

Variant interpretation in clinical genomics requires integration of population-specific allele frequencies, curated database annotations, and phenotype evidence. However, variants annotated as pathogenic or likely pathogenic in reference databases may have different allele frequencies across populations, and electronic medical record (EMR) data may provide useful but incomplete clinical context.

In this study, we used the China Medical University Hospital Genetic Biobank (CMUH-GB) and linked EMRs to evaluate ClinVar-annotated candidate variants in a Taiwanese Han population. Genotyped array variants were filtered by quality control, mapped to ClinVar, and prioritized if annotated as pathogenic or likely pathogenic and observed with an alternative allele frequency greater than 0.0001 in CMUH-GB.

After an updated annotation review, EMR linkage, exclusion of known rare-disease cases or ineligible loci, and retention of variants with clinically relevant EMR phenotypes, 11 candidate variants were analyzed. These variants were located in SCN5A, KCNH2, FBP1, PAH, ACADS, TBX6, BRCA1, LDLR, GP6, and SLC4A11. Several candidate variants showed substantially higher allele frequencies in CMUH-GB and the Taiwan Biobank than reported in some external population datasets. Genotype-phenotype association analyses were performed using additive genetic models with covariate adjustment and Benjamini-Hochberg false discovery rate correction. No interpretable association remained statistically significant after correction. SCN5A rs794728912 models for long QT syndrome and cardiac conduction defects were not estimable because no cases were observed among alternative-allele carriers, resulting in sparse-event separation. A nominal GP6 association with coagulation defects did not remain significant after correction. These findings support population-specific reassessment prioritization of selected ClinVar-annotated variants but do not constitute formal ACMG/AMP reclassification.

Our study highlights the value and limitations of integrating hospital-based genotyping data with EMR-derived phenotypes for ancestry-aware variant interpretation in underrepresented populations.